An algorithm combining age, total prostate-specific antigen (PSA), and percent free PSA to predict prostate cancer: results on 4298 cases

Objectives. To (1) determine if patient age and total prostate-specific antigen (PSA) levels could enhance the ability of percent free PSA to distinguish prostate cancer from benign prostate disease within the 4.0 to 20 ng/mL total PSA range; (2) define the probability of prostate cancer based on patient age, total PSA, and percent free PSA; and (3) define a probability cutoff that distinguishes benign from malignant prostate disease. Methods. The 3773 urologically referred patients with serum PSA values between 4.0 and 20 ng/mL had a sextant biopsy diagnosed as either prostatic carcinoma (1234) or benign prostatic disease (2539) within 60 days of serum specimen collection. We created a logistic regression model, using patient age, total PSA, and percent free PSA, to assign a probability of prostate cancer, and tested the model on an additional data set (525 patients) to calculate sensitivity and specificity. Results. An 18% probability cutoff detected 95% of malignant biopsies and identified 34% of negative biopsies in the validation set. This approach yielded an 11% percentage point increase in specificity over percent free PSA alone. A 20% probability cutoff detected 90% of malignant cases and identified 42% of negative biopsies. Conclusions. A prostate cancer probability based on age, total PSA, and percent free PSA is more effective than percent free PSA alone in differentiating benign prostate disease from prostate cancer. This model may assist physicians and patients regarding the need for biopsy.